Deep Learning-Based Estimation of Myocardial Material Parameters from Cardiac MRI

Area of Science:

• Cardiovascular Imaging
• Biomedical Engineering
• Machine Learning in Medicine

Background:

• Accurate estimation of myocardial material parameters is vital for cardiac biomechanics and computational modeling.
• Traditional methods using inverse finite element (FE) analysis are computationally intensive and slow, hindering clinical use.

Purpose of the Study:

• To develop a rapid and accurate deep learning method for estimating left ventricular myocardial material parameters directly from cardiac magnetic resonance imaging (CMRI).

Main Methods:

• A ResNet18-based deep learning model was trained using finite element method (FEM)-derived parameters from 1288 healthy subjects.
• The model directly estimates myocardial material parameters from routine CMRI data.

Main Results:

• The model achieved high accuracy in healthy subjects with mean absolute errors below 0.0146 for Ca and 0.0139 for Cb (relative errors <5%).
• Performance on a small pathological subset (ARV, HCM) showed higher prediction errors, indicating challenges in diseased tissue modeling.

Conclusions:

• A computationally efficient deep learning framework for myocardial parameter estimation was established, bypassing slow FE optimization.
• Further validation is needed for pathological conditions to enable personalized cardiac modeling and enhance clinical decision-making.